Securing consumer electronics: Fingerprint template generation using DFT for enhanced security

With the widespread adoption of biometric authentication systems in various applications, including consumer electronics, the security of biometric data has become a paramount concern due to its immutable nature. This paper introduces a novel approach for securing user templates in fingerprint-based authentication systems. The proposed technique, which is secure, alignment-free, and non-invertible, leverages the mapping of pair-polar structures of minutiae in a 3D-grid and utilizes Discrete Fourier Transform (DFT). Through extensive analysis using publicly available fingerprint databases from the Fingerprint Verification Competition (FVC) 2002 and 2004, as well as a partial fingerprint database derived from FVC2002 DB1, the effectiveness of the proposed technique is evaluated in terms of revocability, diversity, security, and performance. Comparative analysis of Equal Error Rate (EER) values against state-of-the-art techniques demonstrates the robustness and superiority of the proposed approach.


Introduction
Biometrics leverages unique human traits like facial features, fingerprints, and ear shapes, as well as behavioral attributes like gait and voice, to reliably verify an individual's identity.This technology has found widespread use securing various services, data, and applications against unauthorized access.Biometrics are also integrated into consumer electronics to ensure secure user authentication.Additionally, biometric-based access control systems are implemented in networked devices.For instance, Rahman and Bhattacharya demonstrated biometrics for remote access control of corporate machines, while Corcoran et al. presented a method for access control in home networks for media streaming.Although these systems provide enhanced security over traditional password or token-based approaches in consumer devices, safeguarding the original biometric templates stored on these devices is critical to prevent identity fraud.
If compromised, biometric data in a system leads to a permanent loss of an individual's biometric identity.Stored as templates in databases, these data are vulnerable to attacks, as highlighted by Feng and Jain, Chen et al., and Ross et al., who demonstrated that compromised templates can be exploited to reconstruct original biometrics for unauthorized access.To safeguard these templates, various biometric template protection techniques have been proposed.These techniques fall into two main categories: Biometric cryptosystems, and cancelable biometrics, are two approaches for securing biometric templates.Biometric cryptosystems involve binding or generating cryptographic keys from biometric data, through key-binding or key-generation schemes respectively.Key-binding schemes secure templates by combining them with error correction codewords, while key-generation extracts keys directly from the biometric data.Cancelable biometrics transform biometric feature data using parameters so that compromised templates can be replaced by changing the parameters.
Fingerprints are widely used in biometric systems due to their unique characteristics and ease of acquisition.Nowadays, many consumers electronic devices utilize fingerprint-based biometric systems for access and security purposes.However, storing minutiae, commonly used features in these systems, as user templates in databases can pose risks if compromised, potentially allowing illegitimate access or fingerprint reconstruction.This paper presents a novel alignment-free and non-invertible cancelable technique to generate secure fingerprint templates.These templates prevent reconstruction of the original fingerprint image and unauthorized system access even if compromised.Developing such techniques poses challenges, as Jain et al. outlined essential requirements of revocability, diversity, security, and performance.This work addresses those requirements with the following key contributions:

•
The proposed technique generates a revocable template that does not depend on singular points or alignment, simplifying template revocation in the event of an attack.

•
The fingerprint templates generated through this proposed technique cannot be inverted, thwarting reconstruction of the original fingerprint image even if the template is compromised.This prevents unauthorized access by ensuring the biometric data remains securely encrypted.

•
The design of the paper going ahead is as per the following: Segment II gives a compact outline of relevant strategies concerning biometric layout insurance.In Area III, the proposed procedure is explained upon exhaustively.Following this, Segment IV presents the trial investigation.At last, the paper finishes up with Segment V.

literature review
Different procedures have been proposed in the writing with respect to unique mark layout assurance.A significant number of these strategies are revolved around cancelable biometrics, while others depend on biometric cryptosystems.In biometric cryptosystems, existing strategies are commonly ordered into two classes: key-restricting and key-age plans.Key-binding schemes involve associating discrete cryptographic keys with fingerprint data, often using error correction codewords.These bindings are implemented through methods such as fuzzy commitment and fuzzy vault.In contrast, key-generation schemes directly utilize biometric data to generate the key, sometimes with the inclusion of helper data.The methods described in the literature offer Strong security is provided for user fingerprint data by the techniques outlined in the literature.Nevertheless, there are still significant drawbacks, such as the dependence on single points, non-invertibility, revocability issues, and difficulties with.To address challenges such as aligning gallery and probe fingerprint images and potential loss of discriminative features during transformation, this paper introduces a novel technique for fingerprint template protection The suggested technique is based on mapping pair-polar minutiae structures inside a 3D-grid.It also uses a permutation matrix and the Discrete Fourier Transform (DFT) to improve the security of user templates.This method satisfies all requirements for a fingerprint template protection mechanism, including being alignment-free, non-invertible, and independent of single points.In the section that follows, this strategy will be explained in more detail.

Methodology
The suggested method encrypts the original fingerprint template, which consists of small dots, by use of an irreversible conversion made possible by the Discrete Fourier Transform.Pair-polar minutiae structures are used to ensure an alignment-free secure template by describing local relationships in terms of angles and distances.Figure 1 shows an application of the suggested technique for a consumer electronics (CE) device.First, the device uses its sensor to take a fingerprint impression.Then, it uses the Verifinger SDK (Demo) to extract minute details.The secure fingerprint template is then calculated using the sequence of actions depicted in Figure 1.To further aid in template revocability, a user-specific key (called seed s) is included.

Figure 2
Case study elucidating the pair-polar layout of a minutia point as "mj." Users can supply the key one at a time, or the system can generate a unique key at random for each user.The next subsections will offer further details on each of the processes in the suggested approach

Formation of Pair-Polar Structure
M = {mi : 1 < i ≤ n}, where mi and n stand for the ith minutia point and the total number of minutiae in a fingerprint picture, respectively, may be calculated using the given minutia points.Given that mi = (xi, yi, θi), the position and orientation values of a minutiae point are indicated by (xi, yi) and θi, respectively.
The pair-polar structure may be represented as PP = {pi : 1 ≤ i ≤ n} when each mi is seen of as a reference minutia.Here, pi is equal to pi = {(dij, αij, βij) : 1 ≤ j ≤ n and j /= i}.The numbers dij, αij, and βij are computed in this context using the method outlined in Equation ( 1) in relation to the ith reference minutia (xi,yi,θi) and the j th minutia (xj, yj, θj).
shows an example of a pair-polar structure.The acquired pair-polar structures, with each minutia functioning as a reference point, are then used to create the binary fingerprint template.

Creation of Binary Fingerprint Template
After computing the pair-polar structures (PP) of every minutiae, each structure is individually mapped onto a 3D grid with dimensions d, α, and β for each reference minutia point.Thus, the total number of 3D-grids produced by this mapping equals n, which is the number of minutiae in a fingerprint picture.Grid sizes are given as gd×gα×gβ, with each cell representing the values Gd, Gα, and Gβ along the three dimensions.Gd, Gα, and Gβ are expressed as gd/dmax, gα/αmax, and gβ/βmax, respectively.dmax, αmax, and βmax indicate the highest feasible value for the respective dimension.To map each set pi entry in a 3D grid, divide dij, αij, and βij by Gd, Gα, and Gβ, respectively.
given below.
For each minutia point (j = 1, 2,..., n and j ≠ i), all grid cells are set to zero.The cells corresponding to the given locations (x_j_d, y_j_α, z_j_β) are set to one, and this value is maintained even if several values are assigned to the same place.To construct binary vectors of size 1 × N (where N = gdgαgβ), unfold the obtained 3D-grid for each reference minutia point.

Derivation of the Secure Fingerprint Template
After collecting the binary fingerprint template Bt, security measures are implemented using the N-point Discrete Fourier Transform (DFT), followed by binarization of the DFT's real and imaginary components.The DFT is computed for each binary vector within template Bt using Equation.
In this method, fi(u) is a complex vector carrying the N-point Discrete Fourier Transform (DFT) of the ith binary vector bi of template Bt.F(u) is a matrix of dimension n×N having the N-point DFT of all binary vectors in template Bt.Define fi(u) as fik(u)={real(fik(u))+j imag(fik(u)):k=1,2,...,N}.The safe intermediate binary template is computed by binarizing the real and imaginary sections of the kth value of the complex vector fi(u) separately.Specifically, if real(fik(u))>0 and imag(fik(u))>0, the value in the resultant binary vector is set to 1; otherwise, it is set to 0. This binarization assures that the inverse DFT cannot be performed if the template is compromised.As two bits are calculated for each complex value of fi(u), the resulting binary vector (ri) is 1×2N in size.Binarization is done to all complex vectors in the matrix F(u), resulting in a binary matrix Rt=[r1 r2... rn]T with dimension n×2N.To create the safe user template, the binary vectors are permuted using a random permutation matrix P2N×z (Equation 4), where z < 2N.This permutation provides limitless ways for recovering Rt from the final secure fingerprint template.
Equation ( 4) represents Bft, the final secure fingerprint template.This template is non-invertible, which means that the original information cannot be recovered from the secure user template.Algorithm 1 summarizes these operations concisely.The resulting secure fingerprint template is subsequently saved in the database for future authentication reasons.

Calculation of Similarity Among Protected Templates
We use bit-error computation across binary vectors to determine the similarity of two secure fingerprint templates.This calculation takes place in the converted domain, assuring the total security of the proposed approach.Assume Bfq is a fingerprint query template made up of m minutiae and matching m binary vectors.Additionally, let a template maintained in the database for a fingerprint, consisting of n minutiae points and accompanying n binary vectors, be called Bft.The bit-error between each binary vector biq of the query template and all the binary vectors b1t,b2t,...,bnt of the stored template is used to calculate their similarity.The ultimate similarity score, represented by score, is calculated as follows: Figure 3 The input from database The variables i and k are specified in this context as 1≤i≤n and 1≤k≤m, respectively.The phrase "BitErr(bk, bi)" refers to the bit error between the kth byte vector of the query form and the ith hexadecimal vector of the archived template.
As a result, a high score between two secure patterns being compared denotes a high amount of similarity, whilst a low score indicates a low similarity.

Protecting User Template on Consumer Electronics (CE) Devices
During the registration procedure for an owner's devices, the acquired fingerprint data is converted into a secure template using the aforementioned stages.The generated secure blueprint is then saved on the device.During registration, the device generates a random key (referred as the seed s) for the user and uses it to create the secure template.In the case of an assault, the user is asked to re-register with the device During this procedure, the device generates a new protected template for the user, combining the previous fingerprint data with a freshly created random key.As a result, in the case of an assault, the user only lost connection to the secure template, which prevents retrieval of the unique fingerprint data.This approach assures the entire security of fingerprinting data inside the device, reducing the possibility of identity theft.

Figure 4
The minutiae characteristics from input

Figure 5
The similarity score of the fingerprints

Empirical investigation
The suggested technique was evaluated on four publicly accessible Fingerprint Verification Competition (FVC) databases: FVC2002 DB1, FVC2002 DB2, FVC2004 DB1, and a FVC2004 DB2, as shown in Table .In addition, a partial fingerprint database produced from FVC2002 DB1 is used to demonstrate the applicability of the suggested approach for tiny biometric sensors.To create the partial fingerprint database, tiny patches sized 181 × 181 pixels are extracted, each with at least 10 minutiae.Evaluation is carried out utilizing both single-patch (SinP) and multi-patch (MulP) registration techniques.The SinP technique enrolls a single patch and compares it to a single inquiry patch during verification.In contrast, the MulP approach enrolls many patched from a fingerprint picture and matches them with just one patch after verification.Minutiae from fingerprint pictures are extracted using the Verifinger SDK (Demo).The suggested technique's study includes a variety of elements such as recognition efficacy, revocability, variety, and security, all of which are thoroughly described in the following sections.
To evaluate the recognition performance of the proposed technique, several performance measures have been employed, including False Acceptance Rate (FAR), False Rejection Rate (FRR), Genuine Acceptance Rate (GAR), and Equal Error Rate (EER).The evaluation is conducted using standard 1-vs-1 and FVC verification protocols.
The 1-vs-1 technique selects each subject's two initial fingerprint photos.The FRR is calculated by enrolling the first image and comparing it to the second.In contrast, the FVC protocol enrolls each fingerprint picture of a subject and compares it to all other fingerprint photographs of the same subject to calculate FRR.
To determine the FAR in both rules, each subject's initial fingerprint image is enrolled and contrasted to the first fingerprint picture of every other participant in the database.Each database has 100 participants, each with 8 samples, for a total of 100 true trials in the 1-vs-1 procedure and 2800 in the FVC protocol.4950 imposter comparisons were made for both procedures.
Furthermore, verification is performed under two different attack scenarios: the plain-key attack scenario

Efficacy
In the ideal case of a plain-key assault, the Equal Error Rate (EER) number achieved for all four databases is 0%, displayed in Tables 2 and 3. Furthermore, under the worst-case situation of a stolen-key assault, the suggested solution greatly outperforms previous methods.Additional information on the % EER values acquired in the stolen-key threat case will be provided.Although the suggested technique's efficiency is somewhat inferior to the methods used by Ahn et al. and Li and Hu for the alone database FVC2002 DB2, its mean Equal Error Rate (EER) readings of 5.0% and 4.8% are higher than those of these methods, which are 6.5% and 5.4%, respectively.Similarly, the approach described by Wang and Hu performs marginally better under the FVC protocol.However, it is worth mentioning that the templates generated by the suggested approach are in the form of binary and smaller in size.Furthermore, Wang and Hu did not analyze their strategy on tough fingerprint datasets like FVC2004 DB1 and FVC2004 DB2, but our suggested technique achieves superior results for these databases.

Figure 6
Demonstrates ROC curves from several fingerprint databases for the proposed approach in a stolen key threat scenario Furthermore, the suggested approach is tested on an incomplete fingerprint database using the EER values shown in Table 4 for both scenarios of attack and methods.Surprisingly, a 0% EER is obtained in the plain-key situation, and the findings in the stolen-key assault scenario are quite promising, closely mirroring those obtained on an a full-size fingerprinting image the database, i.e., FVC2002 DB1.This demonstrates the viability of the suggested approach when evaluating incomplete fingerprint scans.Furthermore, Table 5 includes information on the computational aspects, demonstrating the efficacy of the offered approaches.
Figure 7 ROC forms from the incomplete fingerprint database for the suggested approach in a stolen key threat scenario The computationally demanding nature of the proposed approach is assessed in terms of the mean time required for both producing a secure fingerprint pattern and matching.These findings were achieved on a computer with a 4-core CPU running at an optimal clock rate of 3.40 GHz and 8GB of memory.Table 5 shows that the mean time needed for template creation and matching is surprisingly short.This demonstrates the speed of computation of the proposed approach in all data bases, including the incomplete fingerprint database.As a result, it implies appropriateness for use in low-end products.
The suggested technique's effectiveness is further proven by analyzing Receiver Operating Characteristic (ROC) plots, with an emphasis on Area Under the Curve (AUC) values.Ideal values for the AUC vary between 0 to 100, with values near to 100 indicating great system effectiveness.Across all four datasets, the proposed approach achieves AUC values close to 100 under the two different 1-vs-1 and FVC protocols, as shown in Fig. 3(a) and Fig. 3(b).Figure 4 also shows ROC curves produced from trials done on the incomplete fingerprint database.These charts demonstrate the effectiveness of the method.The acquired Area Under the Curve (AUC) numbers, which approach 100%, validate the proposed technique's efficacy when matching incomplete fingerprints.Furthermore, histograms of real and impostor score distributions highlight the significant difference between them.
As shown in Figure 5, real score distributions differ significantly from impostor score patterns for both stolen-key (SK) or different-key (DK) assault situations.Furthermore, statistical studies, such as the Kolmogorov-Smirnov test (KS-test) versus the student's t-test, are used to show the difference between real and impostor score distributions.Usually, an elevated KS-test score suggests that the two input ranges are better separated.
The calculated KS-test values based on the proposed approach are consistently near to one across each of the four fingerprint datasets, as shown in Table 6.Furthermore, a two-sample matched t-test is performed with a significance threshold of 5%.In this test, if the |t-stat| value is greater than the t-critical value, both input distributions are deemed well separated.Table 7 shows the t-test outcomes for the suggested approach using four distinct fingerprint databases.The criteria of adequate differentiation is clearly established.To summarize, all of the foregoing evaluations clearly indicate that the suggested strategy is effective in terms of recognition effectiveness for consumer electronics (CE).Furthermore, the proposed approach may be extended to provide secure access for a variety of systems, especially highperformance computing (HPC) systems.

Reversibility
A template protection solution is revocable if it may substitute corrupted or stolen patterns in the database with fully new secure templates.In our suggested approach, we do this by creating a new private template with the same biometric information but with an alternative transformation key.We test our method's revocability by carrying out revoked template attacks of both Types I and II.
A Type-I attack attempts to access the biometric authentication system by utilizing a compromised template and replacing it in the data base with a new template produced from the same fingerprints photo but with a different key.
In contrast, in a Type-II threat, the substituted new template is constructed with a different fingerprinting sample of the identical subject and a unique key.Our findings, as given in Table 7, suggest that the system does not accept a single corrupted template as legitimate during Type I and Type II assaults.This emphasizes the extremely reversible character of the safe fingerprint templates created with our proposed approach Table 7 The success rates of revoked template attacks on various databases are presented as percentages for the proposed techniques Visually demonstrates the unlikability of the proposed technique using mated and non-mated score distributions

variability
It is an essential component of any fingerprint templates protection system, guaranteeing that secure templates created from an identical fingerprint picture are not connected together.This avoids correlation attacks, in which an attacker tries to deduce unique biometric data by examining several secure templates collected from various fingerprint systems in which a person is registered.To evaluate the variety of templates created using the suggested approach, Gomez-Barrero et al. devised a framework.This framework assesses the score distribution between the mated and non-mated sets of templates.Mated scores compare safe templates created from a single fingerprint collection but using distinct keys, whereas non-mated scores compare secure templates from separate individuals with unique keys.The overlap in the ensembles of mated with non-mated scores, shown in Fig. 6, indicates the lack of resemblance among the reliable templates generated by the suggested approach.Gomez-Barrero et al. have proposed a global metric called Dsys, which gives a thorough assessment of the paraphrasing method's diversity.The diversity metric quantifies variety across secure templates, having values ranging from 0 to 1.A score around 0 shows a large difference within the templates, signifying an elevated level of dissimilarity.
The diversity statistic for secure templates measures how different templates created from a single fingerprint picture are.It is expressed by a number between 0 and 1, with a lower value indicating a greater degree of dissimilarity between the templates.The diversity measure, indicated as Dsys, was determined for multiple databases using the suggested approach, obtaining levels close to 0.001, indicating a significant degree of dissimilarity between the templates.This conclusion demonstrates that the protected templates created by the proposed approach have high variability, as shown by both studies.

Safesty
This feature assures that a secure template generated by a technique is safe in terms of non-invertibility and resistance to various attack scenarios.The suggested approach generates a safe user template that is completely non-invertible, making brute force and cross-match assaults unfeasible.The examination of non-invertibility with the previous attack scenarios is discussed below.
• Non-invertibility: Any secure template generated by the proposed approach is kept in the form of a binary, making it difficult to rebuild the original fingerprint picture regardless of the parameter value s is known.Assume the hacked template Bf t plus the seeds are available to an opponent.In this case, the attacker would require a copy of Bt n× N to reassemble the template.However, the binarization process of F(u) (N-point Discrete Fourier Transformation of Bt n× N) with regard to each row makes recovering the DFT problematic, resulting in multiple alternatives.The damaged template and seed s make it impossible to retrieve the DFT matrices F(u) and Bt n× N.This demonstrates the resilience of the secure template.• Brute force attack: an attacker methodically searches for all feasible possibilities to retrieve the original minutiae information, recreate the fingerprint image from the protected template, or obtain unauthorized access to the biometric system by predicting the final template.However, the concept of non-invertibility shows that retrieving the initial minutiae information is impracticable because to the unlimited options for inverting the Discrete Fourier Transform (DFT) using the safe template.Furthermore, attempted illegal access is not viable using the proposed approach.A quantitative analysis to support this assumption is as follows: Assuming a 3D-grid size of 8 × 16 × 16, wherein gd = 8, gα = 16, and gβ = 16, the binary string after successively expanding the 3D-grid would be N = 2048.Section III-C explains that the final secure template's binary strings will have a size of z, where z < 2N.Assuming z = 2N -4, the number of binary string guesses is 22N -4, or 24092.Therefore, the likelihood of properly forecasting a string of binary characters or n binary numbers is exceedingly insignificant (about 0).Thus, it is obvious that the suggested approach displays resilience against brute force attacks.• Cross-match attack: In this scenario, the attacker attempts to correlate several secured templates acquired through different fingerprint services where a user is enrolled, with the goal of identifying a pattern related to the genuine fingerprint data.However, as shown in Part IV-C, the encrypted templates created using the proposed technique are distinct to one another and reveal no information about the initial fingerprint data.Furthermore, a safe template consists of binary vectors formed during two binarization steps, leaving no obvious clues about the fingerprint image's original minutiae.This emphasizes the impracticality of acquiring significant insights on the first fingerprint template by comparing several secure templates, hence demonstrating the proposed technique's robustness against cross-match attacks.
Fingerprint-based systems for authentication are often used in a variety of uses, including electronic (CE) products.While providing a secure access method, protecting the user's unique fingerprint pattern kept on these devices is a big difficulty.This study presents a unique strategy to preserving a person's fingerprint template that makes use of noninvertible and alignment-free approaches based on pair-polar minutia point structures and the Discrete Fourier Transform (DFT).This approach generates a middle binary fingerprint template by translating pair-polar data onto a 3D grid.The digital template is then safeguarded using DFT and a subsequent binarization method.The resultant output is then permuted with user-specific seeds to create the final secured user template, which is saved in a database for authentication.The suggested approach is evaluated on four distinct biometric records: FVC2002 DB1, FVC2002 DB2, FVC2004 DB1, and FVC2004 DB2, taking into account important aspects such as revocability, diversity, security, and speed.Additionally, assessment is done on an incomplete fingerprint database taken from FVC2002 DB1.A comparative examination with existing methodologies reveals the suggested approach's efficacy and resilience.Future expansions of this technology may investigate its use in multi-biometric systems to generate safe fingerprint templates.

Future scope
Application those provides solutions, support, feedback and problem conclusion.These all are required to be uniquely identifying in future fingerprint can be a most important factor for authentication and authorization.The used of UID to test fingerprint at different places with different application can make it feasible examine the originality of person presented If the Govt.Election may conduct using UID card, then fake entries can be avoided.If he ATM Machine and Card ma connect with the UID card system then only allowed people would transact money Authenticate by Fingerprint Scanner at ATM.Other Scheme those can take advantage to Fingerprint Scanned Images by UID are as follow: Indian Post Office

Figure 1 A
Figure 1 A schematic illustrating the different stages of the proposed method

FVC2004_DB1_TEMPFigure 8
Figure 8 Score distributions for genuine users< imposters with stolen keys, and imposters with different keys across various databases For the purpose of safeguarding fingerprint templates, Jin et al. suggested a kernelized PCA-based method that takes into account several finger samples.A method for blind system identification that quantizes minute pairs to create secure fingerprint templates was presented byWangand Hu.Sandhya et al. created cancelable fingerprint templates by utilizing features from Delaunay triangulation.For the purpose of fingerprint template security, Sandhya and Prasad presented fused bit-strings made from both local and remote minutiae features.By changing minutiae locations, Ali et al. presented cancelable strategies that enhance performance.These techniques are based on the non-invertible transformation of minutiae locations.Using user-specific keys, Trivedi et al. calculated a feature vector by combining the internal angles of Delaunay triangulation with minutiae orientations.A feature-adaptive projection approach was introduced by Yang et al. to secure features that are computed from pairs of tiny points.Lahmidi et al. performed irreversible template creation by grouping minutiae around a single point and computing altered values using user-specific keys, but with performance drawbacks.
These schemes commonly utilize techniques like quantization, fuzzy extractors, or secure sketch methods Drop based methods, as presented by Ratha et al., offer an elective way to deal with address weaknesses in biometric validation frameworks.These methods by and large fall into two classes: salting and non-invertible change.Saltingbased approaches include changing elements utilizing a client explicit key.In such methodologies, the security of the key is significant, as the change is invertible.Alternately, non-invertible change strategies don't need secure keys, as the change utilized is non-invertible.This implies that regardless of whether both the changed format and keys are compromised, recreating the biometric information becomes infeasible.Since the proposed procedure falls under cancelable biometrics, this part gives a survey of different cancelable layout insurance methods The Bio Phasor technique, which Teoh and Ngo introduced, generates a binary code by repeatedly combining the fingerprint data with pseudo-random integers.Cartesian, polar, and functional transformations of biometric features are some of the methods that Ratha et al. developed for computing cancelable templates.An strategy based on

Table 1
Percentage values of EER under the FVC protocol

Table 2
and 3 present a summary of the suggested technique's performance across four datasets.The suggested strategy clearly outperforms previous techniques in the two different 1-vs-1 and FVC protocols, with the sole exception being PMCC64 and PMCC128 .However, it's vital to keep in mind that PMCC64 and PMCC128 missing that revocability attribute, whereas the suggested approach guarantees the production of completely revocable user templates.

Table 3
Average time taken in template generation

Table 4
Average time taken for template generation (FVC database)

Table 5
Average time taken for template matching

Table 6
Average time taken for template matching